What Can Poss(ai)bly Go Wrong? Epoch #01 - Well Sadly, Some Things :/

No pre-reading required!

General Idea: Each series will have 20-30 mins spanning across the various article(s).

Series A

AI Safety Atlas Chapter 2.3: Dangerous Capabilities (https://ai-safety-atlas.com/chapters/v1/risks/dangerous-capabilities/)

Series B:

Anthropic Alignment Faking* (https://www.anthropic.com/research/alignment-faking)

*Sadly this reading is arguably a core one in the AI Safety/Responsible AI domain, afraid there's no choice to choose from for this week!

Agenda

6:30-6:35 (5 mins): Welcome/Intros

--

6:35-6:55 (20 mins): Series A

6:55-7:25 (30 mins): Discussion

--

7:25-7:30 (5 mins): Break

--

7:40-8:00 (20 mins): Series B

8:00-8:30 (30 mins): Discussion

End